Hey all. Anyone tried to measure cpu time of a child worker? All in all I was only able to limit cpu

Hey all. Anyone tried to measure cpu time of a child worker? All in all I was only able to limit cpu time and obviously measure wall time.

Graphql analytics isn’t option for me due to latency, rate limits, etc.

VVlad Lykhonis Hey all. Anyone tried to measure cpu time of a child worker? All in all I was on...

Peps•7/27/25, 11:50 PM

tried using a tail worker?

XxPhi is ok to post job offer here?

Vero•7/28/25, 1:16 PM

No https://discord.com/channels/595317990191398933/775391577790611466/1060602630608519208

PPeps tried using a tail worker?

Vlad LykhonisOP•7/28/25, 2:40 PM

No, but this seems overly complicated for what I want.

What I want is this:

const cpuLimit = 200;

const worker = c.env.appServiceDispatcher.get(
  c.var.serviceUuid,
  {},
  {
    limits: {
      cpuMs: cpuLimit,
    },
  }
);

const startTime = performance.now();
const response = await worker.fetch(request);
const endTime = performance.now();

const duration = Math.min(Math.floor(endTime - startTime), cpuLimit);

const cpuLimit = 200;

const worker = c.env.appServiceDispatcher.get(
  c.var.serviceUuid,
  {},
  {
    limits: {
      cpuMs: cpuLimit,
    },
  }
);

const startTime = performance.now();
const response = await worker.fetch(request);
const endTime = performance.now();

const duration = Math.min(Math.floor(endTime - startTime), cpuLimit);

One of obvious ways is to simply return some sort of CF- header indicating stats back to calling worker/dispatcher?

I do not think creating new worker to observe my customers workers execution is viable here.

VVlad Lykhonis No, but this seems overly complicated for what I want. What I want is this: ```...

Peps•7/28/25, 3:15 PM

not sure i quite understand, because in your example you're basically measuring wall time, not CPU. Any fetches done in the worker would add up to that.
Right now I believe the only solid way of getting the CPU time is by either using a tail worker or logpush. Perhaps in the future we'll get more ways of measuring CPU time

PPeps not sure i quite understand, because in your example you're basically measuring ...

Vlad LykhonisOP•7/28/25, 7:46 PM

Yes I need cpu time not wall time. The best thing I was able to do is measure wall time and cap it at a cpu limit I specify.

It’s quite important for my product, because by definition workers I deployed are my customers code, and I want to charge them by cpu time of their code.

Would not it be possible to simply return some headers back to dispatcher worker? It’s a common pattern in several other products I think?!

Walshy•7/28/25, 7:47 PM

Your best bet would be a tail Worker

Walshy•7/28/25, 7:47 PM

and that seems like the better fit regardless, you'll want likely other stats from these and maybe logs. A Tail Worker will give you all info

WWalshy and that seems like the better fit regardless, you'll want likely other stats fr...

Vlad LykhonisOP•7/28/25, 7:56 PM

Okay I will look into this more. It seems it will cost extra to run it though. It doesn’t scale well considering I’m trying to price CpU time for my customers same as CF does

Maybe I miss understanding tail workers. Will read more.

Walshy•7/28/25, 7:57 PM

Tail Workers do not cost a request, only CPU which if you're just sending off elsewhere will be basically 0 ms

Walshy•7/28/25, 7:57 PM

So it should be extremely cheap

WWalshy Tail Workers do _not_ cost a request, only CPU which if you're just sending off ...

Vlad LykhonisOP•7/28/25, 7:59 PM

Makes sense. Since most of it will be io indeed. This will also scale later for me as we will add observability features. Thank you!

WWalshy Your best bet would be a tail Worker

Vlad LykhonisOP•7/28/25, 8:20 PM

Would you have some pointers on how to get CPU time? I will test out events received tomorrow.

Walshy•7/28/25, 8:36 PM

When you get your event, it should have a cpuTimecpuTime and wallTimewallTime

yourfinance.guy•8/2/25, 10:47 PM

@Walshy | DevPlat EE Hey its Justin with Evafi they said to contact you in here we are really inneed of technical support liek promised

Samko•8/6/25, 7:44 AM

Hello, I have got this email after canceling "Workers for Platforms" but we still have "Workers" active - It's very hard to understand why we got this email will our services work after 19th?

Rafflesia•8/8/25, 8:20 PM

hey! How would I let a user owned domain link to their worker in our dispatch namespace

deleted_user_3434563•8/11/25, 2:10 AM

How well does python run on workers? What are the limitations?

hype•8/11/25, 7:36 AM

How quick is it to deploy a user worker with a couple of dependencies? Is there a way to speed it up?

David•8/16/25, 6:23 AM

Hello i have a question, why did i get ban from wplace?

DDavid Hello i have a question, why did i get ban from wplace?

Hello, I’m Allie!•8/16/25, 6:48 AM

?blocked

SuperHelpflare•8/16/25, 6:48 AM

If you are blocked by a website that uses Cloudflare, you need to contact the website owner to appeal or investigate the block. Cloudflare is configured by the site owner, and Cloudflare will not overwrite the configuration of any customer. Nobody in this Discord nor any Cloudflare employee will be able to help you. The site owner can investigate why you were blocked through your Ray ID (usually found at the bottom of the blocked page).

HHello, I’m Allie!?blocked

David•8/16/25, 6:55 AM

Its all fine thx for the answer ! Ive got a ban message but its all good now

SuperHelpflare•8/17/25, 1:07 PM

Please do not post your question in multiple channels/post it multiple times per the rules at #

welcome-and-rules. It creates confusion for people trying to help you and doesn't get your issue or question solved any faster.

SoBeASH (🥃,🦉, 🔮,🗝)•8/18/25, 6:22 PM

gm - is there a way to move a WfP namespace and everythign in it to another account? I hit the 500 worker limit on my account and then realized I should move to WfP... and then realized that I should really have multiple accounts with WfP in the other account, lol.

Talles•8/19/25, 12:10 AM

guys , can a worker run a linux executable like "./program" ??

TTalles guys , can a worker run a linux executable like "./program" ??

Isaac McFadyen•8/19/25, 1:47 AM

No it cannot.

Isaac McFadyen•8/19/25, 1:47 AM

Workers runs in the V8 sandbox (same one used by Chrome) so it's limited to JS/WASM, or languages that can compile to JS/WASM like Typescript, Rust, etc.

Walshy•8/19/25, 1:47 AM

I mean, could in Containers now

WWalshy I mean, could in Containers now

Isaac McFadyen•8/19/25, 1:48 AM

Oh yeah true!

Isaac McFadyen•8/19/25, 1:48 AM

But it wouldn't strictly be in a Worker, it would be in a container that runs alongside the Worker (not the same level of scale etc), so you'd need to keep that in mind when planning out an application.

Talles•8/19/25, 1:58 AM

@Isaac McFadyen how can I have a proxy on cloudflare?

theo•8/19/25, 7:14 PM

Hi there - I'm looking for a good solution to run some code on behalf of my customers, in such a way that I cannot access their secrets. I actually control all the code, although in the future having client uploaded code might be helpful.

The big issue is I don't want to be able to access their secrets, and ideally I wouldn't be able to arbitrarily update the code which could then let me exfiltrate the secrets. Obviously I'll still want to be able to update the worker code, or deploy a new worker and re-bind the secret, but only with the customer's approval.

Seems like workers for platforms is fairly close to being able to do what I want, but not quite there.

Any ideas on how I can accomplish this?

Ttheo Hi there - I'm looking for a good solution to run some code on behalf of my cust...

Hard@Work•8/19/25, 7:26 PM

Would it work if the account it was deployed to was owned by the Customer?

Hard@Work•8/19/25, 7:27 PM

That's the only thing I can think of. WfP could be an option, but there is no way to lock deploys to be "at your customer's" approval, at least not baked in in a way that you wouldn't be able to change it yourself

HHard@Work Would it work if the account it was deployed to was owned by the Customer?

theo•8/19/25, 8:32 PM

Ideally not. Not all our customers will be necessarily be running on cloudflare, and that would add quite a bit of friction to getting things started.

theo•8/19/25, 8:34 PM

In the past I had posted an idea about a concept of something like a self-managed worker. Where there is some code running within the worker which returns a token, required to update that worker code. So basically you could implement whatever logic you need (some kind of auth, proof of ownership / etc) that gates the update.

We'd still have our users take some manual action to opt into the latest version of our software, which would be available for them to audit and review.

Ttheo In the past I had posted an idea about a concept of something like a self-manage...

Hard@Work•8/19/25, 8:46 PM

I'm not sure I follow your idea there?

You can set it up so that deployments only happen once your system has registered user consent, but there is no way to lock it on the Cloudflare side to prevent you from deploying anyway(at least, not that I'm aware of)

theo•8/19/25, 8:50 PM

Yeah it would have to be something new on the cloudflare side. A way of saying - this worker opts into self-managed updates.
Then CF would have to lock redeploying new code to that worker without some condition being met, likely passing in some kind of token that I as the platform operator dont have access to, but that my users do.

theo•8/19/25, 8:52 PM

If a secret is bound to that worker on creation and never recorded, then you'd have the ability to isolate access from the platform.

LrGian•9/2/25, 3:20 PM

Hello, has anyone in their development process, in their local environment, used a single local d1 for several workers since each worker within .wrangler creates its own d1

rohanbojja•9/2/25, 3:36 PM

Hello, what’s the recommended way to deploy a worker to a dispatch namespace? I noticed the REST endpoint allows uploading scripts, but our use case involves deploying full-stack apps (React + Vite + Hono). Currently, we use Wrangler in our pipeline, would it be possible to achieve the same via the API?

Walshy•9/2/25, 3:41 PM

Yep, you can just do what Wrangler does. The API is the same between Workers & Workers for Platforms, no limitations on what you can upload

You can also use wrangler by specifying --dispatch-namespace--dispatch-namespace in

wrangler deploy

wrangler deploy

Walshy•9/2/25, 3:42 PM

Without Wrangler, you'd want to likely bundle the server side and upload that. For Assets, upload those with the upload API (https://developers.cloudflare.com/workers/static-assets/direct-upload/)

Tamada•9/8/25, 6:27 AM

Hello, is there someone who can help me in a workers that is using Workers for platform? when I started the worker, the message below is showing? please note that app-sandbox-proxy is also running locally.

wrangler 4.20.5
node: v20.19.1

env.DISPATCH_NAMESPACE          Dispatch Namespace      not supported
  app-sandbox-namespace (outbound -> app-sandbox-proxy)

env.DISPATCH_NAMESPACE          Dispatch Namespace      not supported
  app-sandbox-namespace (outbound -> app-sandbox-proxy)

liusha.bnb•9/12/25, 6:31 AM

@Walshy | DevPlat EE hi bro, i'm building a product based on worker for platform, i uploaded several project's static assets as user worker, my question is does these static assets requests counted into billing or not, because i can see dispatch worker logs these static assets requests

Leo•9/12/25, 11:47 AM

There are high packet losses and ping issues on the Germany Telekom line. Is there a solution to this problem?

Lliusha.bnb @Walshy | DevPlat EE hi bro, i'm building a product based on worker for platform...

Isaac McFadyen•9/12/25, 2:56 PM

?pings

SuperHelpflare•9/12/25, 2:56 PM

Please do not ping community members for non-moderation reasons. Doing so will not solve your issue faster and will make people less likely to want to help you.

Hey all. Anyone tried to measure cpu time of a child worker? All in all I was only able to limit cpu

Similar Threads

Similar Threads

Similar Threads